Skip to content
All library documents

Using PCA to Build Currency Baskets and Identify FX Drivers

Article Quant Q&A · Author: Matt Wolf
Read the original

Summary

The document points readers seeking research on currency baskets to two approaches. One cited paper applies principal component analysis to a basket of nine foreign-exchange pairs over a rolling two-month window. It then uses the first principal component as a prediction target, applying logistic regression, decision trees, and neural networks with technical indicators as predictors. This links dimensionality reduction with supervised prediction in a multi-currency trading context.

A second suggestion is to consult the Bank for International Settlements’ currency-index research for methodology on constructing currency indices. The document itself does not reproduce either source’s formulas, results, or implementation details, so it serves mainly as a literature pointer. It also does not compare the suggested approaches or establish that PCA baskets are profitable. The referenced setup is specific to the paper described, and its findings should not be generalized beyond the stated basket and rolling-window design without consulting the underlying research.

Key ideas

  • PCA can reduce information across multiple currency pairs into components that summarize shared movements.
  • The cited study uses the first component as a prediction target.
  • Logistic regression, decision trees, and neural networks are described as prediction methods using technical indicators.
  • BIS currency-index papers are suggested as a source for basket construction methodology.
  • The document provides references rather than evidence of trading performance.

Tags

ForexStatisticsMachine learningTechnical indicators
Full text
# Looking for Research Paper on Creation of Currency Baskets


# Looking for Research Paper on Creation of Currency Baskets












I came across a paper, not sure it originated from academia or a blog or such, that reported on applying principal components to build currency baskets from a set of individual currency pairs and to identify driving currencies. When I use the term basket then I mean a collection of individual cash fx pairs and to relate them linearly or non-linearly through different aggregator functions and weights. I am not talking about basket options here.

Has anyone come across such literature/paper/treatise/blog? I have spent a considerable amount of time in this area and only skimmed through that particular article, thought I had it bookmarked but apparently lost it. I am very interested in other related literature as well.

Thanks

## Answer by Felix (score 3, accepted)

https://quant.stackexchange.com/a/14905

Perhaps this paper by Hyun Woo Byun and coauthors is what you're looking for: Using a Principal Component Analysis to develop Multi-Currency Trading algorithms in the FX market

They apply principal component analysis to a currency basket of 9 pairs with a 2 month rolling window. In a second step, various techniques (logistic regression, decision trees, neural networks) are used to make a predictions of the first PCA component, using technical indicators as predictors.

## Answer by rupweb (score 0)

https://quant.stackexchange.com/a/14229

Try the BIS, they are the experts in this area. http://www.bis.org/statistics/eer/

There are two papers there which explain the methodology behind their currency indices.

Shown in full with attribution under the source's licence. Licence: CC BY-SA 4.0 (Stack Exchange)

This summary was written by Stratmill's research agent from the original; it is not a copy of the source.